University of South Florida researchers have uncovered fundamental flaws in the first US study to claim red light cameras decrease accidents. Since 2001, the insurance industry's report on the benefits of red light camera use in Oxnard, California has been cited by hundreds of cities as the basis for the adoption of photo enforcement (view study in PDF). Researchers Barbara Orban, Etienne Pracht and John T. Large attempted to replicate these findings and discovered that the Oxnard numbers, intended to serve as the model of peer-reviewed scholarship, simply did not add up.

"The regression analysis of [Oxnard study authors Richard] Retting and [Sergey] Kyrychenko does not support their conclusion that red light cameras reduced total or injury crashes," the University of South Florida team wrote in the American Journal of Public Health last month.

In 2004, North Carolina A&T University Professor Mark Burkey was the first to publish a detailed critique of the methodology used in the Oxnard report (read critique in PDF, see page 13). The Florida researchers verified Professor Burkey's findings.

"The Oxnard red light camera study violates many basic principles of sound statistical public health research and lacks internal and external validity," the Florida researchers concluded. "All red light camera investigations should be scrutinized for adherence to applied research methods since studies with greater adherence to quasi-experimental research designs have concluded red light cameras are associated with large increases in crashes and since special interest groups with a financial stake in red light camera use are actively working to influence public opinion and policy."

A number of observers have pointed to conflicts of interest involved in the Oxnard study. The Insurance Institute for Highway Safety funded the research which, in turn, helped its parent companies collect millions in additional profit. Because widespread installation of cameras has increased the number of photo tickets issued in California, each of which carries license points, these companies have been able to collect substantially higher annual insurance premiums. In 2001, the former majority leader of the US House of Representatives slammed the Oxnard study's primary author for not disclosing his own fundamental conflict of interest.

"Before joining the Insurance Institute for Highway Safety, Retting was a top transportation official in New York City at the time the city began looking into becoming the first jurisdiction in the country to install red light cameras," a 2001 report from the Office of the House Majority leader stated (view report). "In other words, the father of the red light camera in America is the same individual offering the 'objective' testimony that they are effective."

As of September 29, Retting was no longer employed by the Insurance Institute. He now works for Sam Schwartz Engineering, a toll road consulting firm.

Article Excerpt:

Analysis Violates Principles of Sound Research & Public Health Evaluation
October 30, 2008
by John T Large, Assistant Professor University of South Florida, Barbara Orban, Etienne Pracht

The study by Retting and Kyrychenko concluding red-light cameras (RLCs) were effective in reducing total and injury crashes [Ref.1], was subsequently reported to be seriously flawed. Burkey and Obeng provide an excellent critique of the deficiencies including the incorrect reporting of results [Ref.2]. Despite this, the Insurance Institute for Highway Safety (IIHS), which is funded by automobile insurance companies, funded the Retting and Kyrychenko study and has continued to use the study in their efforts to advance public policy in favor of RLCs [Ref.3-7]; hence, the importance of ascertaining the study's validity.

The regression analysis of Retting and Kyrychenko [Ref.1] does not support their conclusion that RLCs reduced total or injury crashes. We replicated Retting and Kyrychenko's analysis and verified Burkey and Obeng's critique [Ref.2]. Their regression analysis was simply the use of 16 observational units with 12 dummy variables (df = 3). Retting and Kyrychenko1 attempted to measure red-light camera effectiveness by comparing intersections in a single city that installed cameras (Oxnard, California) with three cities that did not. The variable of interest, labeled "camera," included all 125 signalized intersections in Oxnard, of which 11 had a camera in one approach direction and 114 had no cameras installed. The comparison case in the regression analysis was non-signalized intersections in Oxnard. As such, the "camera" finding merely represents the difference in crude accident growth rates between signalized and non-signalized intersections in Oxnard.

Contrary to the reported results, our calculated difference in the crude growth rates (p = 0.061) was not significant at the 0.05 level. Retting and Kyrychenko incorrectly reported a p-value of 0.0281 for the "camera" variable [Ref.1]. The estimated coefficient (-0.07296) for the "camera" variable in our replicated study matches Retting and Kyrychenko's, but the p-values and degrees of freedom differ. Their rebuttal to Burkey and Obeng was that the signalized/non-signalized variable was statistically insignificant in their initial model and dropped from the model, leaving four degrees of freedom to the error [Ref.8]. However, this was not explained in their methods section and is incorrect since this variable was significant in the initial model (p = 0.00051). Further, when excluding the signalized intersection variable, the coefficient for their "camera" variable is still non- significant (p = 0.2131).

The Oxnard RLC study violates many basic principles of sound statistical public health research and lacks internal and external validity. First, the RLC "treated" intersections were not separately analyzed. Second, the purpose of RLCs is to reduce crashes due to red-light running and yet red-light running crashes were not analyzed. Third, the authors admit to deviating from the methods described. Of most concern, the study's reported statistical results cannot be replicated. All RLC investigations should be scrutinized for adherence to applied research methods since studies with greater adherence to quasi-experimental research designs have concluded RLCs are associated with large increases in crashes [Ref.2,9] and since special interest groups with a financial stake in RLC use are actively working to influence public opinion and policy.

References

1. Retting RA, Kyrychenko SY. Reductions in injury crashes associated with red light camera enforcement in Oxnard, California. Am J Public Health. 2002, 92: 1822-1825.

2. Burkey M Obeng KA. A Detailed investigation of crash risk reduction resulting from red light cameras in small urban areas. July, 2004; Urban Transit Institute. Transportation Institute. North Carolina Agricultural & Technical State University (prepared for U.S. Department of Transportation). Available at: http://www.calccit.org/itsdecision/serv_and_tech/Safety/urbantransitinstitute.pdf. Accessed June 12, 2008.

3. Insurance Institute for Highway Safety. Q&As: Red light cameras. February, 2008. Available at: www.iihs.org/research/qanda/rlr.html. Accessed June 12, 2008.

4. Oesch SL. Statement before the District of Columbia Public Roundtable on automated enforcement: Automated red light and speed camera enforcement. February 23, 2005; Insurance Institute for Highway Safety. Available at: http://www.iihs.org/laws/testimony/pdf/testimony_slo_022305.pdf. Accessed June 12, 2008.

5. Oesch SL. Statement before the Ohio House Committee on Transportation, Public Safety and Homeland Security: Red light camera research. May 4, 2005; Insurance Institute for Highway Safety. Available at: http://www.iihs.org/laws/testimony/pdf/testimony_slo_050405.pdf. Accessed June 12, 2008.

6. McCartt AT. Statement before the Ohio Senate Committee on Highways and Transportation: Red light camera research. October 25, 2005; Insurance Institute for Highway Safety. Available at: http://www.iihs.org/laws/testimony/pdf/testimony_atm_102505.pdf. Accessed June 12, 2008.

7. Oesch SL. Statement before the Pennsylvania House Committee on Transportation: Research on red light cameras. September 27, 2007; Insurance Institutes for Highway Safety. Available at: http://www.iihs.org/laws/testimony/pdf/testimony_slo_092507_rlc.pdf. Accessed June 12, 2008.

8. Kyrychenko SY, Retting RA. Review of "A detailed investigation of crash risk reduction resulting from red light cameras in small urban areas" by M. Burkey and K. Obeng. November, 2004; Arlington VA, Insurance Institute for Highway Safety. Available at: http://www.iihs.org/research/topics/pdf/r1034.pdf. Accessed June 12, 2008.

9. Garber NC, Miller JS, Abel RE, Eslambolchi S, Korukonda S. The impact of red light cameras (photo-red enforcement) on crashes in Virginia. June 2007; Virginia Transportation Research Council. Research Report. Available at: http://www.virginiadot.org/vtrc/main/online_reports/pdf/07-r2.pdf. Accessed February 18, 2008.